On Wed, Jun 11, 2003 at 21:13:52 -0400,
Edmund Dengler <edmundd(at)eSentire(dot)com> wrote:
> Tried the query. Looks like this suffers the same fate as now(): the
> planner does not deal with this as a pseudo constant, so therefore the
> estimator over-estimates the number of possible rows, and thereby chooses
> the wrong operator (a "merge join") to apply.
I tried some more variations and couldn't get it to use an index.
It seems odd that a timestamp without time zone is getting treated
as not constant (presumably because the timezone could change).